search

jhb86253817 / PIPNet

Efficient facial landmark detector
MIT License
412 stars 82 forks source link

Questions about the confidence of key points #35

Closed z411500976 closed 2 years ago

z411500976 commented 2 years ago

Hi,I'm sorry, I still don't understand, the max_cls returned by function forward_pip is the confidence scores of predicted landmarks. why is the confidence of the occluded points higher than that of the unoccluded points? For example, if the face is partially occluded, I want to know which key points are occluded，Or the occluded key points are not displayed. Can it be achieved based on the confidence of key points?

jhb86253817 commented 2 years ago

Hi, I think the scores in max_cls can reflect the confidence of occlusion in general. However, PIPNet may not be suitable for such task, compared to traditional heatmap models, because PIPNet is supposed to be more robust, thus less sensitive to occlusion. A suggestion is that you may try removing the occlusion as data augmentation to make it somehow sensitive to occlusion.

z411500976 commented 2 years ago

thanks again for your reply